Protein Translation and Cell Death: The Role of Rare tRNAs in Biofilm Formation and in Activating Dormant Phage Killer Genes

We discovered previously that the small Escherichia coli proteins Hha (hemolysin expression modulating protein) and the adjacent, poorly-characterized YbaJ are important for biofilm formation; however, their roles have been nebulous. Biofilms are intricate communities in which cell signaling often converts single cells into primitive tissues. Here we show that Hha decreases biofilm formation dramatically by repressing the transcription of rare codon tRNAs which serves to inhibit fimbriae production and by repressing to some extent transcription of fimbrial genes fimA and ihfA. In vivo binding studies show Hha binds to the rare codon tRNAs argU, ileX, ileY, and proL and to two prophage clusters D1P12 and CP4-57. Real-time PCR corroborated that Hha represses argU and proL, and Hha type I fimbriae repression is abolished by the addition of extra copies of argU, ileY, and proL. The repression of transcription of rare codon tRNAs by Hha also leads to cell lysis and biofilm dispersal due to activation of prophage lytic genes rzpD, yfjZ, appY, and alpA and due to induction of ClpP/ClpX proteases which activate toxins by degrading antitoxins. YbaJ serves to mediate the toxicity of Hha. Hence, we have identified that a single protein (Hha) can control biofilm formation by limiting fimbriae production as well as by controlling cell death. The mechanism used by Hha is the control of translation via the availability of rare codon tRNAs which reduces fimbriae production and activates prophage lytic genes. Therefore, Hha acts as a toxin in conjunction with co-transcribed YbaJ (TomB) that attenuates Hha toxicity.


Introduction
Biofilm formation is a complex process that requires the coordinate expression and simultaneous regulation of many genes [1] for reversible and irreversible attachment, microcolony formation, formation of a stable three-dimensional structure, and dispersion [2]. In the absence of conjugation plasmids [3], early steps in biofilm formation require the synthesis of different bacterial surface appendages including flagella that allow reversible attachment [1] and cell motility which is a determinant of biofilm architecture [4]. For irreversible attachment, flagella synthesis is repressed and adhesive organelles like curli fimbriae, encoded by the csg operon, and type I fimbriae, encoded by fim genes, are required for biofilm formation [1]. The mannosesensitive, type I fimbriae also mediate adherence that is important for invasion of host cells in some urinary tract infections [5].
In E. coli, transcription of the type I fimbriae structural gene cluster fimAICDFGH is driven by a single promoter located upstream of fimA [6]. The expression of the fim operon is altered by phase variation that consists of switching an invertible 314 bp element upstream of fimA by two recombinases encoded by fimB and fimE [7]. fimB mediates recombination in both directions (on-to-off and off-to-on) while fimE causes on-to-off recombination [8]. Phase variation of the fimA promoter as well as transcription of the fim genes is influenced by several global, environmental regulators; for example, H-NS reduces the frequency of phase variation [9] and stimulates the transcription of fimA in a locked switched-on mutant [10]. The integration host factor IHF encoded by ihfA and ihfB also plays a dual role in the regulation of type I fimbriae expression since it is required both for inversion of the switching element in both directions and for efficient transcription of fimA [11].
In this work, we investigated the mechanism by which Hha (hemolysin expression modulating protein) controls E. coli biofilm formation. Hha is a small transcriptional regulator (8 kDa) that belongs to the Hha-YmoA family which includes a group of sequence-related, small proteins involved in gene regulation for Gram-negative bacteria [12]. Mechanistic studies of Hhamediated gene expression has focused on its repression of the E. coli a-hemolysin operon hly [12]. However, Hha also represses the expression of many other genes including an endoglucanase of Clostridium cellulolyticum [13], and several virulence genes in pathogenic E. coli [14,15] and Salmonella enterica [16]. Rather than binding directly to specific DNA sequences in the hly operon, Hha exhibits nonspecific DNA binding; therefore, it was proposed Hha binds to H-NS, which binds to specific regions of regulatory sequences in the hly operon [17]. We showed previously that hha is induced 30-fold in E. coli biofilms relative to planktonic cells [18] and that a double deletion of ybaJ-hha affects biofilm formation, motility, and plasmid conjugation [19]. Here we demonstrate that Hha represses biofilm formation by repressing type I fimbriae formation. We also present evidence of a novel, translation-based mechanism for the repression of fimbrial proteins: Hha represses minor regulatory tRNAs that participate in the modulation of the expression of the recombinases FimB and FimE and the structural fimbrial proteins (FimAICDFGH). In addition, Hha exerts a toxic bacteriolytic effect through repression of the rare tRNA; by altering protein translation with the rare tRNAs, general proteases are induced (e.g., ClpP/ClpX) as well as lytic cryptic phage genes. The toxic effects of Hha are counteracted by YbaJ.

Hha represses initial biofilm formation and inhibits cell growth
Since biofilms have a complex time-and temperature-dependent formation process [20], we analyzed biofilm formation of the hha deletion mutant as a function of time, both in LB and LB glucose (LB glu), since Hha has been linked to catabolite repression [21]. In LB glu, the isogenic hha deletion mutant formed an order of magnitude more biofilm initially on a per cell basis than the wild-type strain, and Hha was the cause of this change in biofilm formation since expressing hha in trans reduced biofilm formation to levels even lower than wild type ( Figure 1A). Similarly, in LB, the hha deletion increased biofilm formation 58-fold (at 4 h) but this difference diminished with time ( Figure 1B). As in LB glu, complete inhibition of biofilm formation was observed at 4 h in the complemented hha mutant in LB medium with 2 mM IPTG; complementation also occurred in both media at 0.05 mM concentrations of IPTG (data not shown). Hence, Hha inhibits biofilm formation dramatically and consistently. Note the hha deletion mutation did not affect the growth of planktonic cells in both LB and LB glu media (specific growth rates unchanged).
While overexpressing Hha to complement biofilm formation, we noticed it decreased cell growth in both media as a function of expression level ( Figure 2); hence, overexpression of Hha is toxic. It was verified that the toxic effect observed was a consequence of Hha overexpression and not an artifact created by the addition of IPTG since growth of planktonic BW25113 hha/pCA24N in both LB and LB glu was not significantly decreased by the addition of IPTG (2 mM) (data not shown).

Hha inhibits biofilm formation by repressing type I fimbriae
To elucidate the mechanism by which Hha inhibits biofilm formation, a series of DNA microarrays for biofilm cells cultured in glass wool were performed (4 h in LB and a temporal series of 4 h, 15 h, and 24 h in LB glu for a hha mutant vs. wild-type) since Hha affects the production of multiple proteins and causes pleiotropic effects on catabolite repression and other systems [21]. The biofilm mass formed by the hha mutant in the glass wool cultures was significantly higher than for the wild-type strain which is consistent with the biofilm quantification experiments using crystal violet staining in 96-wells polystyrene plates. Deleting hha impacted the differential expression of a large number of genes (1.3 to 5.5% of the E. coli genome).
In order to simplify the analysis, we focused on those genes that appeared differentially regulated in at least two of the four microarrays (Table S1 and S2). Among the induced genes, stress response genes were well represented as expected [18,20,22,23] including those related to osmolarity (e.g., osmCEY, ybaY, and ygaM) which reflects the role of Hha as an osmoregulator [24]. Additionally, four regulators were induced including the biofilm regulator bssR that regulates biofilm by influencing the uptake and processing of autoinducer-2 (AI-2) as well as indole transport [25] and the carbon storage regulator csrA that regulates the cell central carbon flux, represses biofilm, and activates biofilm dispersal [26]. Also induced were the putative transcriptional regulator yiaG, induced by AI-2-mediator MqsR [27], and ihfA (himA), which together with ihfB (himD), constitutes the global regulator IHF that is required for site-specific recombination, DNA replication, transcriptional control, and fimbriation [28]. fimA, the major constituent of type I fimbriae and an early stage biofilm factor, was also induced upon deletion of hha in early biofilms (4 h) in both LB and LB glu (Table S1). The induction of fimA, ihfA, and ihfB upon deleting hha gave us the first clues that part of the mechanism of Hha inhibition of biofilms is repression of type I fimbriae. Interestingly, ybaJ was significantly induced in all of the microarrays. Since this gene is part of an operon with hha and ybaJ is the first gene downstream of their common promoter, the microarray data indicate that Hha represses ybaJ as well as its own transcription, a common property among the histone-like regulatory proteins like IHF [29], H-NS [30], and HU [31].
The repressed genes upon hha deletion included those related to metabolism, genes with unknown function, the tRNA proM, and fimD, a constituent of the type I fimbriae chaperone usher pathway. fimD repression was another link between Hha and type I fimbriae.
tnaA that encodes tryptophanase, which synthesizes the cell signal indole that is an inhibitor of E. coli biofilm formation [32,33], was induced 2.8-fold at 4 h and then repressed 5-fold at 15 h (LB glu arrays). This significant repression of tnaA at 15 h should decrease indole synthesis and contribute to enhanced biofilm formation for the hha mutant. Corroborating this whole transcriptome data, the extracellular indole concentration of the hha mutant was 1262-fold lower at 15 h in LB glu (435665 mM for the wild-type strain and 2563 mM for the hha mutant) which is consistent with the microarray data.

Confirmation that Hha represses Type I fimbriae via yeast agglutination
To confirm the DNA microarray prediction that Hha represses fimbriae production, a direct evaluation of type I fimbriae production in the hha strain was performed using a classical yeast agglutination assay on glass slides [34] (Figure 3). The results confirm our hypothesis since the hha deletion mutant showed much higher yeast agglutination than the wild-type strain (fimbriae increase agglutination), and the enhanced agglutination was completely abolished by the overexpression of Hha from both pCA24N-hha and pBAD-hha. These results corroborate the DNA microarray data and show that Hha represses the production of type I fimbriae. As expected, the isogenic fimA and ihfA deletion mutants did not show agglutination, and the agglutination phenotype was recovered by the complementation with pCA24N-fimA and pCA24N-ihfA, respectively ( Figure 3).
All cases of positive agglutination were blocked by the addition of mannose 2% which shows that the observed yeast agglutination was indeed caused by the expression of mannose-sensitive type I fimbriae [34]. Additionally, the yeast agglutination phenotype of a hns isogenic mutant was measured since H-NS binds Hha making a heteromeric complex that regulates gene expression [17], and H-NS decreases the frequency of fimA promoter switching [9] as well as activates transcription from the fimA promoter in a locked-on mutant [10]. The results corroborated that H-NS is an activator of fimA expression [10] since no agglutination was observed for the hns deletion mutant under the assayed conditions ( Figure 3); therefore, Hha and H-NS play opposite roles in the regulation of type I fimbriae expression.
Hha partially represses biofilm formation through type I fimbriation via fimA and ihfA Along with fimA, the main structural component of type I fimbriae, hha deletion induced ihfA; ihfA is a subunit of the IHF global regulator that is required for the normal expression of fimA, as mutations in ihfA or in the IHF beta-subunit ihfB (also induced upon hha deletion) prevent the phase variation expression of fimA [11]. Furthermore, when the fim switch, containing the fimA promoter region, is locked in the on phase, lesions in either ihfA or ihfB result in a sevenfold reduction in expression of fimA [11]. Therefore, since the overall effect of IHF in fimA regulation and fimbriae expression is positive, and since both ihfA and fimA are induced upon hha deletion, we hypothesized that another part of the mechanism by which Hha represses biofilm formation is through the diminution of type I fimbriae through the repression of fimA, either directly or indirectly through the repression of IHF.
To test this hypothesis, we evaluated biofilm formation of isogenic fimA and ihfA mutants and found the fimA deletion completely abolished biofilm formation in LB media at 7 h and 24 h and reduced biofilm formation relative to the wild-type strain by 7062% in LB glu at 24 h (data not shown); these results were expected given the known role of type I fimbriae in biofilm production [35]. Moreover, the ihfA mutant had identical biofilm formation as the fimA mutant in both LB and LB glu media since it also completely abolished biofilm formation at 7 and 24 h in LB and repressed by 7162% biofilm formation in LB glu at 24 h (data not shown). This is the first report of the role of IHF in enhancing biofilm formation.
Since the biofilm data were consistent with our hypothesis, biofilm formation of the hha/fimA and hha/ihfA double mutants was evaluated since if induction of fimA via ihfA is important for the increase in biofilm upon hha deletion, then hha deletion in a fimA or ihfA background should not increase biofilm formation to the same extent as the hha deletion in a wild-type background because the production of type I fimbriae should be blocked by the second deletion (fimA or ihfA). As a control, we tested biofilm formation of the double mutants tqsA fimA and bssS fimA since the bssS and tqsA deletions significantly increase biofilm formation in LB glu [25,36], and the mechanism for this increase is not related to the induction of type I fimbriae but instead involves a decrease in indole for the bssS mutant [25] and deficient transport of AI-2 for the tqsA mutant [36]. Hence deleting fimA in these backgrounds should not inhibit the biofilm formation to the same extent as that of a wildtype strain. Corroborating our hypothesis, biofilm formation of the hha fimA and hha ihfA double mutants was completely inhibited in LB at 24 h and 50% inhibited in LB glu at 24 h (showing a similar effect as fimA and ihfA single mutants) (data not shown). In contrast, the bssS fimA and tqsA fimA double mutants still produced a significant amount of biofilm (data not shown). Therefore, Hha represses biofilm formation through type I fimbriae production.
Hha represses transcription of fimA, ihfA, and ybaJ but does not change fimA promoter switching frequency In order to further validate the hypothesis of Hha biofilm inhibition by type I fimbriae diminution through the repression of ihfA and fimA, transcriptional activity of the fimA and ihfA promoters were evaluated in the wild-type strain, the hha mutant, and upon overexpression of Hha from pCA24N-hha. These two promoters as well as the ybaJ-hha promoter (in order to test if Hha repressed its own and ybaJ transcription) were cloned into pPROBE-gfp[tagless] that contains a promoterless, short-lived GFP [37]. As expected, the fluorescence signal derived from the transcriptional activity of the three promoters (fimA, ihfA, and ybaJhha) was higher in the hha mutant than in the wild-type strains ( Figure 4). Also, for all three promoters, the fluorescence was significantly reduced by the overexpression of Hha from pCA24Nhha upon the addition of IPTG 0.5 mM ( Figure 4). Moreover, repression of fimA, ihfA, and ybaJ by Hha was further corroborated by quantitative real time PCR (RT-PCR) using RNA from 15 h LB glu biofilm cells since we found fimA, ihfA, and ybaJ were induced in the hha mutant as expected: fimA 6.160.2-fold, ihfA 3.160.1-fold, and ybaJ 8.060.5-fold. These results corroborate that the activities of the three promoters are repressed either directly or indirectly by Hha.
Since Hha repression of fimA transcription can be a consequence of shifting the fimA promoter orientation to the off position or a result of repression of fimA transcription, or a simultaneous change in both factors (promoter orientation and transcription), we compared the fimA promoter orientation in the wild-type strain and in the hha mutant by PCR-restriction enzyme analysis [38]. No appreciable change was found in the promoter orientation, which was mainly in the on position for both strains (data not shown); hence, Hha represses type I fimbriae transcription but does not affect the fimA promoter orientation significantly.

Hha binds prophage genes and rare tRNA regions in vivo
To probe the whole genome to find which promoters Hha represses and to avoid its nonspecific binding in vitro [17], we performed a nickel-enrichment DNA microarray using biofilm cells. In this assay, formaldehyde is used to cross-link His-tagged Hha to the DNA fragments to which Hha is bound. Then the Hha:DNA cross-linked complexes are purified by a His-tag nickel affinity chromatography and identified using a DNA microarray [39]. Note that H-NS and any other Hha-binding partner were present in the cells during this experiment and thus the identified targets should be physiologically-relevant. For the nickel enrichment DNA microarray assays, biofilm cells were cultured for 24 h at 37uC in LB glu medium, and under these conditions the hha mutant makes 3-fold more biofilm than the wild-type.
The nickel-enrichment DNA microarrays identified 33 genes and 37 intergenic regions enriched at least 4-fold in the hha mutant upon the induction of Hha in trans (Table S3). The intergenic regions and the promoters of the genes enriched represent possible in vivo DNA binding sites for Hha. Among the putative Hha DNA binding sites identified, 15 genes or intergenic regions belonging to the DLP12 and CP4-57 prophage gene clusters were found as well as regions close to argU (encodes an arginine tRNA that recognizes a rare codon), ileY (encodes an isoleucine tRNA that recognizes a rare codon), and rzpD (encodes a predicted murein endopeptidase of DLP12 prophage); these regions are related to Hha toxicity (below). In addition, other putative Hha binding sites were identified such as those for fimbrial-like adhesins, acid resistance, and virulence genes (which supports the role of Hha as a modulator of the expression of virulence factors). We confirmed that Hha regulates the expression of three of the identified genes, ybcW, ypjC, and yeeO, that were enriched 9.2-, 4.6-, and 147-fold, respectively (Table S3), by RT-PCR; using biofilm cells grown as in the nickel-enrichment assay, ybcW, ypjC, and yeeO were induced by 1762, 961, and 5065-fold, respectively. These data confirm Hha binds these chromosomal regions and induces their transcription. Another set of Hha nickelenriched microarrays performed in E. coli BL21 (DE3) by us confirmed that Hha binds to the rare tRNA regions that were identified in BW25113 (argU and ileY), to another two rare codon tRNAs (ileX and proL), and to the hha neighbor gene ybaJ (data not shown). To determine if H-NS influences Hha-DNA interactions, we performed the same experiment in an isogenic hha/hns double mutant and found that the number of putative Hha binding sites decreased to 19 genes and 25 intergenic regions with 26 of these targets common between the two backgrounds and 18 were specific for the hha/hns mutant. Therefore, 44 Hha-DNA interactions were lost in the absence of H-NS (Table S3), showing that H-NS confers some DNA binding specificity to Hha by forming heteromeric complexes.
Hha represses the transcription of rare codon tRNAs (RT-PCR) Given that the nickel-enrichment DNA microarray analysis shows Hha binds near argU which controls gene regulation through rare arginine codons in S. typhimurium [40] as well as binds ileX, ileY, and proL which encode rare codons for E. coli, we hypothesized that Hha inhibits type I fimbriae through the repression of rare codon tRNAs that leads to a diminution of the translation rate of fimbrial proteins. To test this hypothesis, RT-PCR was conducted for argU and proL tRNAs in exponentiallygrowing planktonic cells upon hha overexpression from pCA24Nhha. As expected, the concentrations of argU and proL tRNA in the wild-type strain were higher than in the hha mutant harboring pCA24N-hha (3.260.1-fold for argU and 4.160.2-fold for proL), and the difference increased upon inducing hha by the addition of 0.5 mM of IPTG (5.660.2-fold for argU and 8.260.4-fold for proL). In contrast, Hha induction did not affect the expression of the abundant tRNAs ileV, leuP, and argQ. Hence, Hha specifically represses the transcription of argU and proL rare tRNAs.

E. coli type I fimbriae genes are biased toward rare codons
Since Hha binds and represses the transcription of rare codon tRNA genes, the abundance of rare codons were analyzed in the type I fimbriae regulators fimB and fimE, in fimZ (not known to regulate type I fimbriae in E. coli but it is a fimbriae regulator in Salmonella sp.), and in the type I fimbrial structural gene cluster fimAICDFGH. The rare codons studied were the arginine AGG and AGA codons recognized by argU, the isoleucine ATA codon recognized by ileX and ileY, and the proline CCC codon recognized by proL since Hha binds to these three RNA loci in vivo. Additionally, the abundance of the rare leucine TTG and CTA codons, recognized by leuX and leuW, respectively, and the rare threonine ACA codon, recognized by thrU were also examined.
The sequence analysis showed that the rare arginine AGA codon was present in all the studied genes at least one time and in all cases the frequency was higher than expected according the E. coli codon usage for the complete genome (Table S4). Also, the rare arginine AGG codon was present at a higher frequency than expected in fimZ, fimI, fimC, fimF, and fimG (Table S4). The rare isoleucine ATA codon was also present at a higher frequency than expected in all the genes except fimH and fimA and was especially abundant in fimZ and fimB having six ATA codons each. The rare proline CCC codon was present at a frequency higher than the expected in fimZ, fimB, fimC, fimD, and fimF and was especially abundant in fimD which contained seven CCC codons. Moreover, the rare leucine TTG codon was present at a frequency higher than expected in all genes except fimA, fimE, and fimD, the rare leucine CTA codon was present in fimZ, fimB, fimI, fimF, and fimG, and the rare threonine ACA was present in all genes but fimE, fimI, fimF, and fimH. These results are summarized in Table S4, and a scheme summarizing the regulatory effects of Hha in biofilms, and its role in cell death is presented in Figure 5. Taken together, these results show that a rare codon usage is a characteristic of the regulatory and structural type I fimbriae genes.

Hha represses fimbriae production by repressing rare codons (agglutination)
To show directly that Hha works by repressing the synthesis of rare codons, we added extra copies of rare tRNAs through plasmid pACYC-RIPL during Hha overexpression (from pBAD-hha) during the yeast agglutination assay. The extra copies of rare tRNA counteracted Hha mediated inhibition of yeast agglutination which corroborates that part of the Hha type I fimbriae inhibition mechanism is through the repression of rare coding tRNAs ( Figure 3). Note there was no interference of Hha-mediated inhibition of yeast agglutination with the empty control plasmid pACYC184 ( Figure 3).

Hha toxicity and biofilm formation are attenuated by adding rare codon tRNAs
Given that Hha is toxic ( Figure 2) and represses the transcription of argU and proL, we hypothesized that part of the toxicity mechanism may involve depletion of rare codon tRNAs since the diminution of specific rare codon tRNAs by minigenes is toxic with the rare arginine AGG and AGA codons (argU) and the rare isoleucine ATA codon (ileX and ileY) most toxic [41]. In order to test this hypothesis, overexpression of Hha was assayed with extra copies of ileY, argU, leuW, and proL using pACYC-RIPL; the addition of additional copies of rare codon tRNAs through pACYC-RIPL significantly decreased Hha toxicity ( Figure 6A). Furthermore, the addition of additional copies of rare codon tRNAs through pACYC-RIPL significantly increased biofilm formation of BW25113 hha/pBAD-hha (0.3560.09 vs. 0.0960.05) ( Figure 1B). Hence Hha toxicity, fimbriae production, and biofilm formation are linked to the rare codons.

YbaJ attenuates Hha toxicity
Based in the following observations, our data suggest that Hha and YbaJ may form a toxin-antitoxin module with Hha the toxin ( Figure 2) and the poorly-characterized YbaJ the antitoxin: (i) hha and ybaJ are in the same operon, (ii) the separation between the genes is short (25 bp), (iii) both proteins are small, (iv) Hha exerts negative transcriptional self-regulation, and (v) overexpression of Hha is toxic; all of these are characteristics of a toxin-antitoxin pair [42]. Further supporting this hypothesis is that the ybaJ-hha operon Figure 6. Extra copies of rare codon tRNAs and YbaJ reduce Hha toxicity. Inhibition of cell growth in LB at 37uC by expressing Hha from plasmid pET28-hha with 0.5 mM IPTG in E. coli BL21 (DE3)/pACYC184 (empty vector control) or in BL21 (DE3)/pACYC-RIPL cells expressing extra copies of the rare codon tRNAs ileY, argU, leuW and proL (A). Effect of YbaJ overexpression from pBAD-ybaJ in E. coli K-12 BW25113 hha for Hhamediated toxicity from pVLT31-hha in LB at 37uC; Hha was induced by the addition of 2 mM IPTG and YbaJ was induced by the addition of 0.1% Larabinose (B). The experiments were performed in duplicate, and one standard deviation is shown. doi:10.1371/journal.pone.0002394.g006 is highly induced in E. coli biofilms [18], and toxin-antitoxin modules arise at a higher frequency in stationary growth phase and in biofilms in association with persister cells [43].
In order to test if YbaJ counteracts the toxic effects of Hha, we assayed the effect of YbaJ overexpression in Hha toxicity using a two plasmid system consisting of pBAD-ybaJ, to induce YbaJ by the addition of 0.1% arabinose, and pVLT31-hha, to induce Hha by the addition of 2 mM IPTG. As expected, the induction of Hha without YbaJ was toxic to the cell, and the toxicity of Hha was attenuated by simultaneous YbaJ overexpression ( Figure 6B). Hence, YbaJ expression diminished Hha toxicity.

Hha causes cell lysis by inducing prophage lytic genes and by activating toxins via proteases
Given that Hha binds argU and ileY in vivo (Table S3) and that Hha toxicity in part is mediated by depletion of rare codon tRNAs ( Figure 6A), the gene context of these tRNAs was examined. Two of the tRNAs that Hha binds, argU and ileY, are close to prophage gene clusters since 37 DLPD-12 prophage genes are located immediately downstream to argU. We have found 10 of these DLPD-12 prophage genes are differentially-regulated in biofilms relative to planktonic cells [23,36]. Furthermore, 26 CP4-57 prophage genes are located five genes upstream ileY, and we have found eight of these genes are differentially regulated in biofilms [23,36]. Of these 18 prophage genes related to biofilms, 10 are significantly repressed in hha mutant biofilms at 15 h and 24 h in LB glu (Table S6), and no DLPD-12 or CP4-57 prophage gene is induced in the absence of Hha at any time in any condition; therefore, it appears Hha either directly or indirectly activates the expression of some phage genes that cause cell lysis (probably by repression of argU and ileY). Specifically, upon hha deletion, prophage gene essD is repressed 2.6-fold at 24 h in LB glu; EssD encodes a putative holin [44] and forms an operon with ybcS and rzpD (http://ecocyc.org/) that encode a putative endolysin [44] and a predicted murein endopeptidase (http://ecocyc.org/) respectively. alpA is also repressed 4.9-fold at 15 h in LB glu and encodes a transcriptional activator of a P4-like cryptic prophage [45]. Furthermore, yfjZ-ypjF is repressed 2.4-fold at 24 h in LB glu and 6-fold at 15 h LB glu, respectively, and encodes a toxinantitoxin module [46]. Corroborating the transcriptome microarray data and the nickel-enrichment DNA microarray data, induction of Hha in the rzpD, yfjZ, alpA, and appY deletion mutants resulted in significantly less growth inhibition than in the wild-type hha mutant background ( Figure S1) which indicates that these DLP12 and CP4-57 prophage genes are involved in Hha toxicity. In contrast, Hha overexpression did not have a significantly different effect on the growth of the other prophage mutants like ybcY, ybcN, ypjF, yfjF, yfjG, ybcS, essD, and ompT (data not shown).
To investigate further the hypothesis that Hha overexpression activates lytic prophage genes leading to cell death, formation of lysis plaques in soft agar upon Hha overexpression from pCA24Nhha by the addition of 1 mM IPTG was assayed; surprisingly, we found that Hha overexpression causes the formation of cell plaques ( Figure S2) and that the treatment of BW25113 wild type with the lysate extracted from those plaques does not cause lysis by itself. This data agree with the lysis seen in Figure 2B upon induction of Hha with 2 mM IPTG. Hence, Hha toxicity leads to cell lysis and may not involve the production of active phage, however we can not rule out this possibility.
To explore further the toxicity mechanism of Hha, a DNA microarray was performed for Hha overexpression using BW25113/pCA24N-hha suspension cells grown at 37uC with 2 mM IPTG to induce hha. Hha induced 39 genes and repressed 5 genes more than 4-fold (Tables S7A and S7B). As expected, hha was induced 6.5-fold, a level of induction comparable to the 30fold induction of hha in biofilms [18]; hence, the induction levels of hha in our experiments are not far from those naturally observed and the effects of Hha overexpression then are physiologically relevant. Of those induced genes, three encoded the proteases ClpP, ClpX and Lon which degrade many proteins including the antitoxins of several toxin-antitoxin pairs [47]; therefore, it is likely that induction of these proteases by Hha degrades antitoxins leaving the toxins free to exert inhibitory effects on growth. Supporting our hypothesis, overexpressing Hha in both clpX and clpP deletion mutants resulted in significantly less growth inhibition than in the wild-type hha mutant background ( Figure S1E). Moreover, three antitoxin genes were also induced: dinJ, relB and yefM, which may be a response of the cell to compensate for the low antitoxin protein levels, and RelB and YefM are known targets of Lon [47]. In addition, 12 heat shock proteins (mainly chaperones) were induced upon Hha overexpression which is reasonable in that Lon degrades RelB which stimulates relBE transcription [48] causing RelE to accumulate in excess which causes translation inhibition and perhaps the accumulation of misfolded proteins that trigger the activation of the heat shock chaperones.
The Hha overexpression microarray data were validated by RT-PCR since RT-PCR showed lon, dinJ, and yefM were induced 2.560.4-fold, 4.660.4-fold, and 3.360.3-fold, respectively. Additionally, induction of Hha in the relE and yoeB toxin deletion mutants resulted in significantly less toxicity than in the hha mutant (data not shown) corroborating that Hha expression induces other toxins and that these toxins are required for Hha toxicity. Taken together, these results suggest that an important part of Hha toxicity may be indirect activation of several toxins of toxin-antitoxin pairs via the degradation of their antitoxins by regulatory proteases.

YbaJ DNA-binding regions
To research the role of YbaJ in the attenuation of Hha toxicity, nickel enriched DNA microarrays were also performed using pCA24N-ybaJ to identify putative in vivo binding sites for YbaJ in biofilms cells. The results indicate that YbaJ may bind to 21 genes and two intergenic regions (Table S3). Of these putative targets, the more interesting include the ybaJ gene itself, indicating that YbaJ may regulate hha and its own expression. Also, YbaJ binds sfmH, which encodes a predicted Salmonella fimbrial-like adhesin protein that is only one gene away from fimZ and argU (Hha binds these sites), which indicates Hha and YbaJ bind nearby loci and may compete for the regulation of genes near the binding sites. Since Hha exerts a toxic effect counteracted by YbaJ, YbaJ binding ybcS, which encodes the prophage endolysin gene of the DLP12 prophage cluster genes that lies near argU and which is repressed in hha mutant biofilms, indicates YbaJ may repress this gene as part of its mechanism. Additionally, YbaJ binds ariR (ymgB), which encodes an acid resistance gene involved in the regulation of biofilm formation and which we found to be a structural Hha analog [39].

Hha induces biofilm dispersal
Since bacterial cell death produced by bacteriophage [49] and the activation of killing proteins [50] promotes biofilm dispersal in pseudomonads, we explored the possibility that Hha could cause dispersal in biofilms. To test this hypothesis, biofilm cells of E. coli K12 BW25113/pCA24N-hha/pCM18 were cultured in LB glu at 37uC in a continuous flow cell system for 42 h to form a biofilm, then 1 mM IPTG was added to induce Hha expression for 6 h. We found that Hha expression reduced significantly the biomass, substratum coverage, and average thickness by 28620%, 38620%, and 27616%, respectively, compared to the biofilm before IPTG addition. In contrast, when IPTG was not added, biomass, substratum coverage, and average thickness increased by 32620%, 32619%, and 28618%, respectively at 48 h. Therefore, Hha inhibits biofilm formation and contributes to the detachment of preexisting biofilms.

Discussion
Since Hha was discovered as a repressor of episomal hemolysin production [51], it has been studied extensively to discern its regulatory role in the cell. Hha affects the production of multiple proteins and causes pleiotropic effects for genes related to catabolite repression [21]; Hha also affects other phenotypes including plasmid supercoiling [52] and insertion sequence transposition [53]. Hha is conserved in pathogens [53] and also represses the expression of other virulence genes in pathogenic E. coli including episomal Vir adhesion [14] and the locus of enterocyte effacement [15].
Here we investigated the relationship between Hha and biofilm formation in E. coli K12 in an attempt to explore its role in cellular physiology, a role that remains unclear despite considerable investigation [12]. Previously we found that the ybaJ-hha operon was induced significantly in E. coli biofilms [18] and that Hha and YbaJ regulate biofilm formation in cells harboring the derepressed conjugative plasmid R1drd19 [19]. Through our temporal study, we show here that Hha controls many genes in both LB and in LB glu including the main structural subunits of the cellular appendage required for attachment, type I fimbriae (fimA). This implies Hha is important for early steps in biofilm formation. In agreement with this, the hha deletion had the most effect in LB at early times ( Figure 1B). Other lines of evidence found here linking Hha to fimbriae include (ii) ihfA encoding a subunit of the IHF regulator that is an activator of type I fimbriae production via the fimA promoter [11] is activated in all the LB glu microarrays upon hha deletion (Table S1), (iii) Hha represses fimA and ihfA transcription in biofilm cells as shown by RT-PCR, (iv) fimA and ihfA mutations abolish biofilm formation, (v) deleting hha along with either fimA and ihfA (double mutants) fails to increase biofilm formation, (vi) transcription of fimA and ihfA is induced in the hha deletion mutant and is inhibited by overexpressing hha (GFP reporter constructs) (Figure 4), and (vii) type I fimbriae production dramatically increases in the absence of hha and Hha overexpression inhibits type I fimbriation ( Figure 3). Together these results show Hha represses fimbriae synthesis although it does not affect fimA promoter orientation.
Although the regular DNA microarrays give valuable information about how Hha controls E. coli biofilm formation (Tables S1, S2, and S6), it is not possible to determine which effects were mediated directly by Hha transcriptional control. However, an intrinsic difficulty for the identification of the promoters which are regulated directly by Hha in vivo is that in vitro approaches like electrophoretic mobility shift assays are not useful for the identification of Hha binding sites since Hha forms heteromeric complexes with other proteins like H-NS that exhibit different DNA binding properties, and Hha alone exhibits nonspecific DNA binding [17]. Therefore, we utilized an in vivo nickel enrichment DNA microarray assay to identify DNA binding based on methods such as nickel agarose-based chromatin enrichment [54]. We found that in vivo, Hha binds its own promoter; this is consistent with induction of ybaJ in all the regular hha DNA microarrays (corroborated with RT-PCR in biofilm cells) and with the enhancement of transcription from the ybaJ-hha promoter observed in the hha mutant ( Figure 4). These experiments demonstrate that Hha regulates its own transcription via negative feedback.
Furthermore, the nickel enrichment DNA microarrays show Hha binds in vivo to the rare tRNA loci argU, ileX, ileY, and proL as it reduces fimbriae production. Salmonella possesses a fimbrial cluster whose regulation is different than that in E. coli since in Salmonella there is no phase variation control by promoter switching. The Salmonella cluster is composed of fimZYW and fimU, and fimU is a tRNA homolog of argU [55,56]. fimU tRNA has an important role in the regulation of type I fimbriae in Salmonella since rare arginine AGA codons are prevalent in the three fim regulatory genes (fimZYW), and production of FimY and type I fimbriae are reduced in the absence of fimU since the fimU mutation results in the inhibition of efficient FimY translation [57]. Other examples of the regulatory effects of tRNA on fimbriae are that the leucine tRNA leuX that recognizes the rare leucine codon TTG controls the expression of the fimB recombinase in uropathogenic E. coli [58], and argU controls the expression of fimY in Salmonella. Both regulatory tRNAs exert their effect by the attenuation of the translational rate of FimB and FimY since these genes are biased for argU and leuX codons [57,58].
The repression of the argU, ileX, ileY, and proL rare tRNAs by Hha should be physiologically relevant for the translation of proteins with a high proportion of rare AGG, AGA, ATA, and CCC rare codons, respectively. Our sequence analysis of fimZ, fimB, fimE, and the type I fimbrial gene cluster fimAICDFGH show all the genes contain an abundance of rare codons recognized by argU, ileX, ileY, proL, as well as rare threonine and leucine codons recognized by thrU and leuXW, respectively (Table S4). Given that both fimEand fimB-promoted switching are stimulated by the amino acids alanine, isoleucine, leucine, and valine [59] and that the use of minor codons is a general mechanism to control the temporal expression of specialized genes at the post-transcriptional level [60], we propose that Hha inhibits the production of type I fimbriae not only by the transcriptional inhibition of ihfA and fimA but also by the diminution of the availability of argU, ileX, ileY, and proL tRNAs and this leads to a reduction in the translation of the type I fimbriae proteins. This hypothesis was corroborated since the addition of extra copies of the genes for the argU, ileY, leuW, and proL tRNAs via constitutive expression in pACYC-RIPL blocked Hha repression of type I fimbriae as demonstrated by increased yeast agglutination (Figure 3).
There are implications for this mechanism for the Hha control of virulence genes since usually they contain a high proportion of rare codons due to their acquisition by horizontal transfer [61]. Hence, Hha-regulated genes may contain an abundance of rare codons so that they may be repressed by Hha at the translational level by the mechanism we described for the fimbrial genes. In support of this, hilA, which controls the virulence genes of the pathogenicity island SPI-1 of S. enterica, is repressed by Hha [16], and our analysis shows it contains an abundance of rare codons including isoleucine ATA (14-fold more than the expected), arginine AGA (31-fold), proline CCC (7-fold), and leucine TTG (6-fold). Therefore, the Hha-rare tRNA translational regulatory mechanism may be important for the regulation of this gene. Another example is that efficient expression of hemolysin encoded by hlyA in uropathogenic E. coli is dependent on the availability of leuX [62], and we observed that hlyA has rare codons for argU (5fold more than expected AGA and 6.6-fold more AGG), for ileX and ileY (3-fold more ATA), for proL (1.53-fold more CCC), and for thrU (2.5-fold more ACA). Since Hha is a well-known transcriptional hemolysin repressor, it is also possible that the diminution of the availability of these regulatory tRNAs also plays an important role in Hha-mediated hemolysin repression. In addition it was found recently in Salmonella that Hha and its paralog YdgT modulate the transcription mainly of horizontally-acquired genes [63]; hence, Hha may be a key regulator of virulence factors at both the transcriptional and translational level as is shown for type I fimbriae genes in this study.
Our data also suggest that Hha may be part of a toxin-antitoxin pair with YbaJ ( Figure 6B); this hypothesis could be confirmed by evaluating Hha-YbaJ protein-protein interactions. One of our main findings is that Hha toxicity stems from cell lysis ( Figure S1), and that at least part of the toxic Hha effect is caused by the depletion of regulatory tRNAs encoding rare codons ( Figure 6A). Our data also indicate that the mechanism by which Hha is toxic may be due to activation of the DLP12 and CP4-57 prophage lytic genes that are nearby the Hha binding sites argU and ileY (Table  S6), and also by the indirect activation of several toxins by the degradation of their corresponding antitoxins by Lon, ClpP, and ClpX proteases and that the rare codon tRNAs repressed by Hha may trigger the activation of these lytic genes and proteases. Hence, cryptic phage genes may be retained by the cell for lysis and are regulated in part by Hha.
One of the most surprising results was that Hha overexpression triggers the formation of lytic plaques; however, the detailed mechanism is a subject for further research. So far we have many promising links between phage activation and Hha; for example, some of the prophage genes that Hha regulates are related to Hha toxicity and some of those genes, like alpA and ihfA, have a role in phage activation (alpA is associated with the excision of CP4-57 [45], and IHF is essential for the lysogenic life cycle of lambda phage [64]). It is possible that ssrA (a tmRNA part of the CP4-57 prophage locus that regulates the expression of many phage) may be under the control of Hha (like the rare codon tRNAs) and both are involved in the activation of phage and cell lysis. Additionally, stress conditions, like starvation, trigger the activation of the lambda phage lytic cycle; hence, the stress produced by Hha overexpression may activate the production of active phage.
YbaJ may exert part of its attenuation of Hha toxicity by repressing these prophage lytic genes since it binds near lytic genes (e.g., sfmH) and in lytic genes (e.g., ybcS) (Table S3). Note biofilm formation by the ybaJ strain in LB glu medium after 24 h was the same as the hha mutant (data not shown); hence, deleting ybaJ likely inactivates hha and leads to the observed increase in biofilm formation. Due its important induction in biofilms and its role in counteracting Hha induced toxicity (that may be mediated by the expression of multiple phage lytic genes and toxins like essD, ybcS, rzpD and ypjF), we propose ybaJ should be named tomB (toxin overexpression modulator in biofilms).
The regulation of cell lysis in biofilms by Hha-YbaJ may be an important aspect of the E. coli biofilm physiology since in the biofilms of Pseudomonas aeruginosa and Pseudoalteromonas tunicate, cell lysis is either mediated by the activation of prophage genes or by autolytic proteins, and these events promote biofilm dispersion and enhance phenotypic variation that may be important for the successful adaptation and colonization to new niches [49]. Therefore, in E. coli biofilms, cell death mediated by Hha may also influence these phenotypes. Furthermore, our results suggest another important function of toxin-antitoxin pairs in bacterial physiology in addition to the nine already proposed [42]: these proteins may also induce biofilm dispersal in response to the environmental conditions. The activation of prophage genes for this cell lysis, as controlled by the availability of rare tRNAs, may have additional biological relevance since biofilms are exposed to nutrient (amino acid) starvation which would lead to the diminution of charged rare tRNAs, and this may activate the prophage killing genes allowing the starving cells to commit suicide which would diminish competition for nutrients with nearby cells, provide nutrients to other cells, promote biofilm dispersal (which also is induced by starvation in Pseudomonas putida [65]), and promote phenotypic variation. A toxin-antitoxin module controlling this process may be a good strategy for the cells to trigger cellular death in response to starvation, since upon starvation, protein synthesis will diminish (including the synthesis of the toxinantitoxin proteins) and since the toxin half-life is greater than the antitoxin half-life, after protein synthesis is diminished, the toxin is free to cause cell death. The fact that Hha expression promotes dispersal via toxicity and via repression of fimbriae may be the physiological reason that explains the apparent paradox of induction of Hha, a biofilm inhibitor, in biofilms, since its function may be to promote the dispersal of those biofilms and to allow the detached cells to stay unattached to promote colonization in distant niches.

Bacterial strains and growth media
The E. coli strains and plasmids are listed in Table S5. E. coli K-12 BW25113, its single isogenic mutants, E. coli AG1, and pCA24N and its derivatives were obtained from the Genome Analysis Project in Japan [66,67]. hha, fimA, and ihfA in the pCA24N constructs were tightly-regulated with the lac promoter and its repressor lacI q by the addition of isopropyl-b-D-thiogalactopyranoside (IPTG, Sigma, St. Louis, Mo.). The promoter-probe vector plasmid pPROBE-gfp[tagless] was provided by Dr. Lindow [37] and was used to construct pPfimA-gfp containing the fimA promoter-gfp fusion, pPihfA-gfp containing the ihfA promoter-gfp fusion, and pPybaJ-gfp containing the ybaJ-hha promoter-gfp fusion. The expression vector plasmid pBAD-Myc-His C (Invitrogen, Carlsbad, CA) was used to construct pBAD-hha and pBAD-ybaJ. Expression of Hha or YbaJ proteins under the arabinose-inducible araBAD promoter (P BAD ) was induced by 0.1% L-arabinose (Acros Organics, Morris Plains, NJ). The broad host range, IPTG-inducible plasmid pVLT31 was provided by Dr. de Lorenzo [68] and was used to construct pVLT31-hha. pACYC-RIPL, a plasmid carrying the tRNA genes argU, ileY, leuW, and proL, was isolated from BL21-CodonPlus cells (Stratagene, La Jolla, CA).
All experiments were conducted at 37uC. Luria-Bertani medium (LB) [69] was used to pre-culture all the E. coli strains with appropriate antibiotics to maintain plasmids. LB and LB supplemented with 0.2% (wt/vol) glucose (LB glu) were used for the crystal-violet biofilm experiments. LB was also used for the growth rate assay, GFP fluorescence intensity assays, and the yeast agglutination assay. LB and LB glu were used for the glass wool biofilm DNA microarrays, and LB glu was used for the nickelenrichment DNA microarrays, for the indole determination assay, and for the biofilm dispersal experiment with flow cells. Kanamycin (50 mg/mL) was used for pre-culturing all of the E. coli BW25113 isogenic mutants, and kanamycin (100 mg/mL), chloramphenicol (30 mg/mL), ampicillin (100 mg/mL), and tetracycline (20 mg/mL) were used to maintain plasmids (for twoplasmid systems, multiple antibiotics were used at the same concentrations). Cells were pre-cultured overnight at 37uC from single colonies with shaking (250 rpm) for the various assays and experiments (biofilm, indole, motility, planktonic growth, Hha toxicity, cell lysis, biofilm DNA microarrays, and nickel-enrichment DNA microarrays). The specific growth rates were measured using cell turbidity at 600 nm using two independent cultures (turbidity less than 0.7).

Plasmid construction
To construct the fimA promoter-gfp fusion, a 585 bp DNA fragment containing the fimA promoter region (from -563 bp to -1 bp upstream of the start codon of fimA) including the 314 bp switching element and the two inverted repeat sequences IRF and IRR was obtained by PCR amplification using BW25113 genomic DNA as template with primers 59-CCCCGAGCTCGCTGCTTTC CTT-TCAAAAAA-39 and 59-CCCCGGATCCCCGCCAGTAATG-CTGCTCGT-39 that form BamHI and SacI restriction sites flanking the fimA promoter. Similarly, to construct the ihfA promoter-gfp fusion, a 314 bp DNA fragment containing the ihfA promoter region (from -296 bp to -4 bp upstream of the start codon of ihfA) was obtained by PCR with primers 59-CCCCGGATCCGATT-TCTCGCTTCCCGGCGA-39 and 59-CCCCGAGCTCGTT-CAATCCCTCAATGATGC-39. Similarly, to construct the ybaJhha promoter-gfp fusion, a 220 bp DNA fragment containing the ybaJ-hha promoter region (from -198 bp to -1 bp upstream of the start codon of ybaJ) was obtained by PCR with primers 59-CCCCGGATCCGCTGC GGGTTAGTGCTAGTA-39 and 59-CCCCGAGCTCGAACGCGTCCCCTTCTTAGCG-39. After double digestion with BamHI and SacI, the DNA fragments were cloned into the promoter-probe vector plasmid pPROBE-gfp[tagless] [37] to obtain the pPfimA-gfp, pPihfA-gfp, and pPybaJ-gfp.
To construct the pBAD-hha plasmid, a 237 bp DNA fragment containing the hha gene (from +4 bp to +216 bp downstream of the start codon of hha) was obtained by PCR amplification using BW25113 genomic DNA as the template with primers 59-CCCCCTCGAGCTCCGAAAAACCTTTAAC GAA-39 and 59-CCCCTTCGAATTAGCGAATAAATTTCCATAC-39 that form XhoI and HindIII restriction sites flanking the hha gene. Similarly, to construct pBAD-ybaJ, a 393 bp DNA fragment containing the ybaJ gene (from +4 to +372 bp downstream of the start codon of ybaJ) was obtained by PCR amplification using BW25113 genomic DNA as template with primers 59-CCCCCTCGAGCGATGAATACTCACCC AAAAG-39 and 59-CCCCTTCGAACTAACAAGATAAACTCGCAG-39 that form XhoI and HindIII restriction sites flanking the ybaJ gene. After double digestion with XhoI and HindIII, the DNA fragments were cloned into pBAD-Myc-His C to obtain pBAD-hha and pBAD-ybaJ, which expresses Hha and YbaJ, respectively, under the L-arabinose-inducible P BAD promoter.
To construct pVLT31-hha plasmid, a 248 bp DNA fragment containing the hha gene (from -5 bp upstream to +222 bp downstream of the start codon of hha) was obtained by PCR amplification using BW25113 genomic DNA as the template with primers 59-GCGAATTCGAAGTATGTCCG-39 and 59-GCGCTCTAGATATTTATTAGCG-39 that form EcoRI and XbaI restriction sites flanking the hha gene.

Crystal-violet biofilm and indole assays
The biofilm assay was conducted in 96-well polystyrene plates [35]. Bacteria were inoculated at an initial turbidity at 600 nm of 0.05 in all media at 37uC or 30uC for 4 h, 7 h, 15 h, and 24 h without shaking, then the growth (turbidity at 620 nm) and total biofilm were measured using crystal violet staining. IPTG addition (2 mM) had no effect on biofilm formation of the wild type strain in both LB and LB glu media. Each data point was averaged from at least twenty replicate wells using at least four independent cultures of each strain. Extracellular indole concentrations of BW25113 and the hha mutant cultured in LB glu were measured spectrophotometrically as described [25] at 15 h. Two independent cultures were used for each strain.

Yeast agglutination assay
The evaluation of type I fimbriae production was measured via yeast agglutination as described [34]. Briefly, bacteria were grown for 24 h at 37uC on LB plates with appropriate antibiotics to select plasmids. For the induction of complemented genes (hha, fimA, or ihfA) from pCA24N, 0.25 mM IPTG was added, and for the induction of hha from pBAD-hha, 1% L-arabinose was added. After growth, cells were collected and suspended in phosphatebuffered saline (PBS, NaCl 8 g/L, KCl 0.2 g/L, Na 2 HPO 4 1.44 g/L, KH 2 PO 4 0.24 g/L, pH 7.4) [69] to a turbidity of 5 at 620 nm and mixed with a 10% w/v brewers yeast suspension (Acros Organics, Morris Plains, NJ). A few drops of the mixture were poured onto microscope glass slides and the yeast agglutination was evaluated after 5 min. Each experiment was performed two times using two independent cultures.

Hha toxicity
The effect of Hha on cell growth was assayed by inducing Hha expression with 0 to 2 mM IPTG using pCA24N-hha. The effect of adding extra copies of the rare codon tRNAs ileY, argU, leuW and proL on toxicity was assayed using BL21 (DE3) codon plus-RIPL/ pET28A-hha which contains rare codon tRNA encoded on pACYC-RIPL; 2 mM of IPTG was added to cells in LB+kanamycin (50 mg/mL)+chloramphenicol (30 mg/mL) to induce Hha expression from pET28A-hha. The effect of YbaJ on Hhamediated toxicity was assayed in BW25113 hha by the simultaneous expression of Hha and YbaJ with 2 mM IPTG using pVLT31-hha to induce Hha and 0.1% arabinose with pBAD-ybaJ to induce YbaJ. The effect of Hha on the cell growth of the prophage mutants was assayed by inducing Hha expression from pCA24N-hha with 1 mM IPTG in LB at 37uC.

Cell lysis plaques
Samples (1 mL) of overnight cultures (cultured with or without 0.5 mM IPTG) were mixed into a soft agar overlay containing 3 mL of media containing 10 g/L tryptone, 1 g/L yeast extract, 8 g/L NaCl, 8 g/L agar, the appropriate antibiotics, and 1 mM IPTG. The overlay was poured onto plates with the same composition but with 10 g/L agar, and the plates were incubated at 37uC for 24 h.

GFP fluorescence intensity assay
To determine the specific intensity of the GFP fluorescence from the pPfimA-gfp, pPihfA-gfp, and pPybaJ-gfp plasmids, the bacterial cultures were grown at 37uC with 250 rpm shaking until late exponential growth phase (turbidity at 600 nm of 1) except for those cultures that contained pCA24N or pCA24N-hha that were grown to turbidity of 0.5 at 600 nm and then 0.5 mM IPTG was added and the incubation was continued for two hours. Aliquots were diluted rapidly with PBS to turbidity at 600 nm of 0.1 to 0.2, and the fluorescence was quantified (Spectra Max Germini EM, Molecular Device, Sunnyvale, CA, USA) using excitation at 485 nm and emission at 528 nm with a 515 nm cutoff. For normalization, the final cell density was determined by measuring the turbidity at 600 nm (Tecan, Austria Gesellschaft, Salzburg, Austria). The GFP negative strain BW25113 pPROBE-gfp [tagless] was used as the negative fluorescence control. All experiments were performed at least twice.

Flow cell biofilm experiments and image analysis
LB glu supplemented with erythromycin (300 mg/mL) to maintain the constitutive green fluorescent protein plasmid pCM18 [72] and chloramphenicol (30 mg/mL) to maintain pCA24N-hha were used to form biofilms at 37uC in a continuousflow cell as described previously [27]. The biofilm was visualized at 42 h and at 48 h (after 6 h of the IPTG addition) with a TCS SP2 scanning confocal laser microscope (Leica Microsystems, Heidelberg, Germany) with a 40 X N PLAN L dry objective with a correction collar and a numerical aperture of 0.55. Confocal flow cell images were analyzed with COMSTAT image-processing software [73] as described previously [74].

RNA isolation and DNA microarrays
Biofilm cells on glass wool for each strain for the four sets of DNA microarrays were prepared as described previously [20]. Suspension cells of BW25113/pCA24N and BW25113/ pCA24N-hha were cultured in 25 mL of LB at 37uC supplemented with 2 mM IPTG (initial turbidity 0.1 at 600 nm) to a turbidity at 600 nm of 0.5. Total RNA was isolated from cells as described [18]. The E. coli Genechip antisense genome array chip (P/N 900381, Affymetrix, Santa Clara, CA) was used to analyze the complete E. coli transcriptome of biofilm cells, and the E. coli Genechip Genome 2.0 array chip (P/N 900550, Affymetrix, Santa Clara, CA) was used to analyze the suspension cells as described [20]; note that there were no comparisons of data from different Genechips. Background values, noise values, and scaling factors of all arrays were examined and were comparable. The intensities of polyadenosine RNA controls were used to monitor the labeling process. As expected, signals of the deleted genes, araA and rhaA, were low for all strains, while the signal of hha was low for strains with hha deleted and high for strains were hha was overexpressed. For both sets of binary microarray comparisons to determine differential genes expression, if the gene with the larger transcription rate did not have a consistent transcription rate based on the 11-15 probe pairs (pvalue less than 0.05), these genes were discarded. A gene was considered differentially expressed when the p-value and the corrected p-value based on the False Discovery Rate Method [75] for comparing two chips were lower than 0.05 (to assure that the change in gene expression was statistically significant and that false positives arise less than 5%) and when the expression ratio was higher than the standard deviation for all of the E. coli K-12 genes (1.4 for the 4 h LB, 1.4 for the 4 h LB glu, 3.3 for the 15 h LB glu, 1.6 for the 24 h LB glu, and 1.8 for the suspension cell arrays) [76].

Microarray accession numbers
The expression data for biofilm samples of the E. coli wild type and hha mutant, and the expression data of suspension cells upon Hha overexpression have been deposited in the NCBI Gene Expression Omnibus (GEO) [77] and are accessible as GSE8708 (for biofilm cells) and GSE10116 (for Hha overexpression with suspension cells).

Nickel-enrichment DNA microarrays
These experiments were performed as reported previously [39]. Briefly, overnight cultures of E. coli K-12 BW25113 hha/ pCA24N (control) and BW25113 hha/pCA24N-hha, BW25113 hha hns/pCA24N-hha, or AG1/pCA24N-ybaJ (expressing Hha or YbaJ with a His 6 -tag at the amino terminus) [67] were diluted into 250 mL of LB glu media with chloramphenicol (30 mg/mL) to maintain pCA24N, pCA24N-hha, and pCA24N-ybaJ and with 1 mM IPTG to induce Hha or YbaJ synthesis. Cultures were grown at 37uC with 250 rpm shaking in 1 L capacity Erlenmeyer flasks containing 10 g glass wool to produce biofilms [18]. After 24 h, formaldehyde (1%, Fisher Scientific Co., Pittsburgh, PA) was added to crosslink the His-tagged Hha or YbaJ proteins and the DNA to which it was bound (present also were proteins like H-NS that may interact with Hha in vivo). Ni-NTA agarose gel resin was used to bind the His 6 -tagged Hha-DNA or His 6 -tagged YbaJ-DNA complexes from the biofilms cells, and RNase A and proteinase K were used to remove RNA and protein. The DNA fragments were labeled [18], and the DNA microarrays were performed as indicated above. Positive candidates for the Hha or YbaJ binding in vivo were those genes or intergenic regions with at least a 4-fold higher signal than the global average signal of all the genes and IG regions in the Hha or YbaJ chips (,600), and those that were at least 15-fold enriched with respect to the signal in the control array (the pCA24N vector). This cut off ratio was selected because the standard deviation of the enrichment (Hha or YbaJ signal divided by empty vector signal) was 1.3 for the Hha experiment and 5.3 for the YbaJ experiment. The spiked control rhaD gene gave a signal from ,13,000 to ,32,000 in the experiments (,22-fold to ,54-fold higher than the average signal) which validated our methodology. Figure S1 Inhibition of growth in LB at 37uC by Hha overexpression via pCA24N-hha is less severe in various mutants: (A) rzpD, (B) yfjZ, (C) alpA, (D) appY, and (E) clpP/clpX. Each experiment was performed three times, and one representative experiment is shown. Found at: doi:10.1371/journal.pone.0002394.s001 (1.81 MB TIF) Figure S2 Plaques in soft agar containing 1 mM IPTG upon expression of Hha at 37uC in BW25113 hha/pCA24N-hha. BW25113 hha/pCA24N was used as the negative control and the effect of 0.5 mM IPTG to induce Hha in the overnight culture is shown. The experiment was performed in duplicate, and one representative experiment is shown.

Table S3
List of gene regions bound in vivo by Hha and YbaJ in biofilms cells at 37uC in LB glu after 24 h as identified by nickelenrichment DNA microarrays in E. coli K12 BW25113 hha, E. coli K12 BW25113 hha/hns, and E. coli K12 AG1. Signal ratio indicates fold change of pCA24N-hha or pCA24N-ybaJ cells relative to cells with the empty plasmid pCA24N. 1 Gene or intergenic region close to argU. 2 Gene or intergenic region close to ileY. 3 Gene or intergenic region close to appY. 4 Prophage gene or intergenic regions next to prophage genes. Found at: doi:10.1371/journal.pone.0002394.s005 (0.12 MB DOC)

Table S4
Rare codons present in E. coli fim genes. The numbers in bold indicate more rare codons than expected according the E. coli codon usage for the complete genome. The expected use of rare codons in E. coli is as follows: 7% of all the isoleucine codons should be ATA, 13% of all the proline codons should be CCC, 4% of all the arginine codons should be AGA, 2% of all the arginine codons should be AGG, 13% of all the leucine codons should TTG, 4% of all the leucine codons should be CTA, and 13% of all the threonine codons should be ACA. The codon usage data were obtained from the Genomic Atlas Database (http://www.cbs.dtu. dk/services/GenomeAtlas/). Found at: doi:10.1371/journal.pone.0002394.s006 (0.07 MB DOC)